Research map / Continual learning / Class-incremental
DDGR: Continual Learning with Deep Diffusion-based Generative Replay
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use diffusion-based generator with classifier-derived instruction-operator (DDGR) instead of GAN/VAE generative replay baselines | continual learning generative replay · CIFAR-100 CI (NC=5, AlexNet) | final average accuracy: +5.99 over best baseline (PASS) | Helped |
| Diffusion generator with instruction-operator (DDGR) vs. baselines | continual learning generative replay · ImageNet CI (NC=50, AlexNet) | final average accuracy: +9.92 over best baseline | Helped |
| Diffusion generator with instruction-operator (DDGR) | continual learning generative replay · ImageNet CI (NC=50, AlexNet) | forgetting rate: 17.31 -> 6.98 | Helped |
| Diffusion generator with instruction-operator (DDGR) | continual learning generative replay · CIFAR-100 CI (NC=5, AlexNet) | forgetting rate: 24.08 -> 23.00 | Helped |
| Diffusion generator with instruction-operator (DDGR) | continual learning generative replay · CORe50 CIR (ResNet) | final accuracy acc_78,FTS: 7.66 higher than best baseline | Helped |
| Reusing generated samples to update the generator (DGR) vs using original previous data (ablation) | DGR · CORe50 preliminary | sample quality (qualitative) | Hurt |
| Diffusion-based DDGR vs. DDGR without the classifier instruction-operator (ablation) | DDGR · CI and CIR | accuracy | No clear effect |
Closest papers by meaning
- SDDGR: Stable Diffusion-based Deep Generative Replay for Class Incremental Object Detection (CVPR 2024 · similarity 0.89)
- Revisiting Generative Replay for Class Incremental Object Detection (CVPR 2025 · similarity 0.86)
- Dynamic Expansion Diffusion Learning for Lifelong Generative Modelling (AAAI 2025 · similarity 0.86)
- Continual Unsupervised Generative Modelling via Online Optimal Transport (AAAI 2025 · similarity 0.86)
- Continual Learning in Modern Hopfield Networks with an Application to Diffusion Models (NeurIPS 2026 · similarity 0.86)
- Overcoming Catastrophic Forgetting in Visual Continual Learning with Reinforcement Fine-Tuning (NeurIPS 2026 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.